معرفی
Antoine Miech is a Researcher at DeepMind's Vision Group, with prior affiliations at Inria and Ecole Normale Supérieure where he completed his computer vision Ph.D. under Ivan Laptev and Josef Sivic. He has collaborated with researchers from Facebook AI and Google during his academic career.
Research Interests span video understanding, weakly-supervised machine learning, and multimodal analysis. His work focuses on:
- Text-video embedding
- Self-supervised video representation
- Action localization
- Anticipatory video modeling
- Scalable multimodal learning
Scientific Contributions include:
- HowTo100M - A massive dataset of narrated instructional videos
- MIL-NCE - A novel loss function for video-text alignment
- MEE - A model for handling heterogeneous data
- Context Gating - Learnable pooling architecture
Awards & Recognition: Google Ph.D. Fellowship (2018)
Technical Leadership: Created the LOUPE TensorFlow toolbox for feature pooling and maintained annotated video dataset catalogs. Organized the Data Science Game competition (2016-2017).
۰مقاله منتشرشده
